Visual Sample Plan and Prior Information: What do we need to know to find UXO?

Kenneth A. Flagg

Montana State University
April 11, 2016
Megan Higgs, Advisor

Introduction and Background

Introduction

  • Military activities leave debris behind
    • Including explosive munitions
  • Hazardous sites should be remediated
  • Statistics can reduce costs
  • Visual Sample Plan (VSP Development Team 2016)
    • Software popular with DoD contractors
  • My project:
    • Understand how sample plan inputs affect VSP mapping
    • Simulation study

Unexploded Ordnance

Situation

Unexploded Ordnance

Boom

Unexploded Ordnance

Boom

Targets of Interest vs Anomalies

  • Target of interest (TOI) – munitions-related item which could be unexploded ordnance (UXO)
  • Anomaly – any item detected by a metal detector

Spatial Anomaly Density

Background + TOI

Visual Sample Plan Overview

Target Area Identification Features

  • Systematic transect sampling plans
  • Analyze sample data
    • Moving average anomaly density
    • Semivariogram models
    • Ordinary Kriging
    • Delineate high-density regions

Parallel Transect Sampling Plan

VSP screenshot showing a systematic transect sampling plan

Anomaly Density Map

VSP screenshot showing estimated anomaly density and high
density regions

What inputs affect the delineation the most?

Prior Info Experiment

  • Simulation Study
    • Three sites
    • Varying complexity: easy, medium, hard
    • Sampling plans created with VSP
    • Kriging and delineation similar to VSP

 

  • Factors:
    • VSP input TA size
      • Levels: Too Small, TA1, TA2, Too Large
    • VSP input TA anomaly density
      • 100/acre, 200/acre, 400/acre

Easy Site

100 realizations analyzed with each spacing

Input TA Size Input TA Anomaly Density
100/acre 200/acre 400/acre
Too Small (849 ft by 566 ft, 8.66 acres) 40 ft 100 ft 220 ft
TA1 (1,200 ft by 800 ft, 17.3 acres) 125 ft 225 ft 465 ft
TA2 (2,000 ft by 900 ft, 32.4 acres) 170 ft 390 ft 655 ft
Too Large (2,828 ft by 1,273 ft, 64.9 acres) 270 ft 565 ft 935 ft

Easy Site

Results

Results

Results

Results

Conclusions

Conclusions

  • Most TOI items detected
  • Too much area delineated
  • Need accurate info about TA size
    • Density info relatively unimportant
  • More data reduces variability but not bias

Conclusions

  • My recommendations:
    • Focus on TA size
    • Use smallest TA for sampling plan
    • Do not sample more than necessary

Acknowledgements

Acknowledgements

  • Megan Higgs
  • Katie Catlett, Stephanie Fitchett
  • Lenovo
  • My Mom
  • YOU!

Appendix

Medium Site

Hard Site

References

References

Matzke, Brett, John Wilson, Lisa Newburn, S.T. Dowson, John Hathaway, L.H. Sego, L.M. Bramer, and Brent Pulsipher. 2014. Visual Sample Plan Version 7.0 User’s Guide. Richland, Washington: Pacific Northwest National Laboratory. http://vsp.pnnl.gov/docs/PNNL-23211.pdf.

Neptune and Company, Inc. 2008. “Helena Valley Probabilistic Risk Assessment of MEC for the Montana Army National Guard.”

Pulpisher, Brent, John Hathaway, Brett Matzke, John Wilson, and Lisa Newburn. 2014. Demonstration Report for Visual Sample Plan Remedial Investigation (VSP-RI) Sampling Methods at the Motlow Site in Tennessee. Richland, Washington: Pacific Northwest National Laboratory. http://www.pnnl.gov/main/publications/external/technical_reports/PNNL-23210.pdf.

VSP Development Team. 2016. Visual Sample Plan: A Tool for Design and Analysis of Environmental Sampling. Richland, Washington: Pacific Northwest National Laboratory. http://vsp.pnnl.gov/.